The Missing Layer for Personal AI | First Person Labs’ Ken Liu
The Missing Layer for Personal AI | First Person Labs’ Ken Liu
Podcast33 min 57 sec
Listen to Episode
Note: AI-generated summary based on third-party content. Not financial advice. Read more.
Quick Insights
  • Treat Second and other personal-AI products as a sector to monitor, not an investable opportunity yet: the company is private, and the discussion provides no valuation or business metrics.
  • Watch NVIDIA (NVDA) as a thematic beneficiary of confidential AI computing, but the episode offers no stock-specific forecast or price target.
  • Focus on the broader opportunity in lower-cost AI inference and privacy infrastructure; falling model costs could enable always-on personal AI, though security, sensitive-data access, and operating expenses remain key risks.
Detailed Analysis

First Person Labs / Second (Private)

  • First Person Labs is developing Second, a personal-AI product that builds a portable model of a user’s preferences, goals, relationships, and digital activity.
    • The product is designed to share that context with other AI tools, so users need not repeatedly explain their preferences or ongoing work.
    • The speakers described uses including remembering development context, carrying work between devices, and surfacing useful reminders without being prompted.
  • The company’s proposed model is a persistent, trusted layer for personal context, while other agents specialize in tasks such as shopping or coding.
  • The transcript describes the product as available, but also says the company is still scaling inference capacity and working to roll it out responsibly.

Takeaways

  • The discussion points to an opportunity in personal AI infrastructure: persistent context, data portability, and coordination across AI tools could become valuable as people use more agents.
  • The business case is not yet established in the transcript. The speakers cite high inference costs, and the product’s usefulness depends on users being comfortable granting access to sensitive information.
  • Security is a central risk: the guest says prompt injection cannot be definitively solved and that some protections rely on model reasoning, alongside deterministic controls.
  • Treat this as an early-stage, private-company opportunity to monitor rather than a publicly traded investment. The episode gives no revenue, valuation, or investment terms.

OpenAI (Private)

  • OpenAI is mentioned through Codex and its models. The guest says Second could provide Codex with work context already gathered elsewhere.
  • The discussion also says OpenAI had recently offered a confidential inference option, which aims to protect data while models process it.

Takeaways

  • The episode suggests two potential areas of competition and investment interest: AI coding tools and privacy-preserving model services.
  • Portable user context could make AI tools more useful, but it may also make it easier for users to switch between providers.
  • The transcript provides no specific recommendation, price target, or financial performance information. OpenAI is not publicly traded.

Anthropic (Private)

  • Anthropic is referenced through its Opus model and pricing. In a hypothetical example, the guest estimates that running ten agents continuously at 60 tokens per second using Opus would cost about $1,000 per day.
  • The example illustrates how continuous background analysis can be expensive, even if it is not a quoted price for a typical user.

Takeaways

  • The discussion highlights the importance of inference cost and efficiency for companies building always-on AI products.
  • A model provider’s ability to deliver capable models economically could matter to adoption, but the transcript does not compare providers’ overall commercial performance.
  • Anthropic is not publicly traded, and the stated cost is an illustrative scenario rather than a price forecast.

DeepSeek (Private)

  • DeepSeek Flash is used as a lower-cost comparison in the same hypothetical example: the guest estimates about $60 per day, compared with about $1,000 per day using Opus under the stated assumptions.
  • The guest also says token prices had fallen by more than 50% since First Person Labs began, and expects costs to keep declining as new models become available.

Takeaways

  • The discussion favors the broad theme of lower-cost, capable models: falling inference costs could make persistent AI services more practical.
  • The transcript does not establish that DeepSeek Flash is equivalent to Opus for every task. The guest says different tasks may call for different models.
  • DeepSeek is not publicly traded; the cost comparison is a specific hypothetical, not a general pricing guarantee.

NVIDIA (NVDA)

  • The guest describes NVIDIA hardware as one possible trust anchor for confidential inference: a model can run in a protected environment that is intended to prevent others from viewing the data being processed.
  • The speakers distinguish this hardware-based approach from zero data retention, which they describe as a provider promise or policy.

Takeaways

  • The episode points to a potential investment theme in confidential computing: AI handling sensitive personal data may increase demand for hardware and systems that can protect inference.
  • The proposed security chain still depends on hardware, verification, and implementation. The transcript does not quantify how much demand this could generate for NVIDIA.
  • This is a thematic observation, not a specific stock recommendation.

Apple (AAPL)

  • Apple is mentioned as an operating-system provider whose systems expose certain activity signals, but the guest says operating systems can also restrict how applications access or use them.
  • The discussion compares local data storage with cloud approaches and mentions iCloud as an example of a familiar synchronization model.

Takeaways

  • The episode highlights the strategic role of device and operating-system control in personal AI: platform permissions may shape which products can access context and deliver always-available services.
  • The same platform control may be a constraint for independent AI developers. The transcript does not discuss Apple’s financial outlook or make a stock recommendation.

Microsoft (MSFT)

  • Microsoft is mentioned alongside Apple and Linux as an operating-system platform that can provide computer activity signals to applications.
  • The guest’s product concept relies on access to a user’s digital activity, while the speakers note that operating systems may limit what applications can do.

Takeaways

  • The discussion underscores the importance of platform access and permissions for personal-AI products.
  • The potential opportunity is indirect: AI products may depend on operating-system integrations, while platform restrictions could limit their functionality.
  • No Microsoft-specific investment recommendation, price target, or financial outlook is given.

Personal AI, Inference, and Privacy Infrastructure

  • The episode’s broader investment themes are personal AI, portable context across AI tools, background agents, model efficiency, and privacy-preserving inference.
  • The guest describes a trade-off between the depth of a user model and its cost: more background analysis may improve understanding but require more inference.
  • The speakers identify specific challenges, including high operating costs, sensitive-data access, security architecture, prompt injection, and operating-system restrictions.

Takeaways

  • Watch for products that can demonstrate both useful personalization and credible data controls; the episode presents trust as essential to adoption.
  • Falling model costs may expand the market for always-on AI, but the transcript makes clear that costs remain material for intensive use.
  • No cryptocurrency, crypto asset, or specific price target is mentioned.
Ask about this postAnswers are grounded in this post's content.
Episode Description
AI agents are getting smarter, but every new agent still has to learn who you are. What if your preferences, priorities, relationships, and digital context could move with you instead? First Person Labs co-founder Ken Liu joins a16z’s Martin Casado and Joel de la Garza to introduce Second, a personal context layer designed to understand what you’re doing across your digital life and make that context available to the agents you choose. They discuss how Second can proactively connect information across conversations and applications, why protecting such personal data requires a different security architecture, and the longer-term vision of AI that doesn’t just respond to prompts, but understands enough context to know when and how to help.  Resources: Try Second: https://2.computer Learn more about First Person Labs: https://fpl.company/ Follow Ken Liu: ​​​​https://www.linkedin.com/in/kenziyuliu/ Follow Martin Casado on X: https://x.com/martin_casado a16z.com Follow Joel de la Garza:  https://x.com/YODELAGARZA a16z.com Stay Updated: Find a16z on YouTube: YouTube Find a16z on X Find a16z on LinkedIn Listen to the a16z Show on Spotify Listen to the a16z Show on Apple Podcasts Follow our host: https://twitter.com/eriktorenberg Please note that the content here is for informational purposes only; should NOT be taken as legal, business, tax, or investment advice or be used to evaluate any investment or security; and is not directed at any investors or potential investors in any a16z fund. a16z and its affiliates may maintain investments in the companies discussed. For more details please see a16z.com/disclosures. Hosted by Simplecast, an AdsWizz company. See pcm.adswizz.com for information about our collection and use of personal data for advertising.
About The a16z Show
The a16z Show

The a16z Show

By Andreessen Horowitz

The a16z Podcast discusses tech and culture trends, news, and the future – especially as ‘software eats the world’. It features industry experts, business leaders, and other interesting thinkers and voices from around the world. This podcast is produced by Andreessen Horowitz (aka “a16z”), a Silicon Valley-based venture capital firm. Multiple episodes are released every week; visit a16z.com for more details and to sign up for our newsletters and other content as well!